Some other covid19 visualizations:

https://coronavirus.1point3acres.com/

https://coronavirus.jhu.edu/map.html

# data source https://www.census.gov/data/datasets/time-series/demo/popest/2010s-state-total.html and wikipedia
df_population <- data.frame(
  state = c("AK", "AL", "AR", "AS", "AZ", "CA", "CO", "CT", "DC", "DE", "FL", 
            "GA", "GU", "HI", "IA", "ID", "IL", "IN", "KS", "KY", "LA", "MA", 
            "MD", "ME", "MI", "MN", "MO", "MP", "MS", "MT", "NC", "ND", "NE", 
            "NH", "NJ", "NM", "NV", "NY", "OH", "OK", "OR", "PA", "PR", "RI", 
            "SC", "SD", "TN", "TX", "UT", "VA", "VI", "VT", "WA", "WI", "WV", "WY"),
  population = c(731545, 4903185, 3017804, 55465 , 7278717, 39512223, 5758736, 3565287, 705749, 973764, 21477737,
                 10617423, 165768, 1415872, 3155070, 1787065, 12671821, 6732219, 2913314, 4467673, 4648794, 6892503, 
                 6045680, 1344212,  9986857, 5639632, 6137428, 56882, 2976149, 1068778, 10488084, 762062, 1934408,
                 1359711, 8882190, 2096829, 3080156, 19453561, 11689100, 3956971, 4217737, 12801989, 3193694, 1059361,
                 5148714, 884659, 6829174, 28995881, 3205958, 8535519, 106977 , 623989, 7614893, 5822434, 1792147, 578759)
)

# The Atlantic Monthly Group (CC BY-NC 4.0)
# source: https://covidtracking.com/api

df_states <- fread("https://covidtracking.com/api/v1/states/daily.csv") %>% 
               replace(is.na(.), 0) %>%
               inner_join(df_population, by = "state")%>%
               mutate(date = as.Date(as.character(date), "%Y%m%d"))

tableau10 <- as.list(ggthemes_data[["tableau"]][["color-palettes"]][["regular"]][[1]][,2])$value
first_day <- as.Date("2020-03-15") # to select a date
today <-  as.Date(toString(max(df_states$date)))
  
kable(head(df_states, n = 3))
date state positive probableCases negative pending totalTestResults hospitalizedCurrently hospitalizedCumulative inIcuCurrently inIcuCumulative onVentilatorCurrently onVentilatorCumulative recovered dataQualityGrade lastUpdateEt dateModified checkTimeEt death hospitalized dateChecked totalTestsViral positiveTestsViral negativeTestsViral positiveCasesViral deathConfirmed deathProbable totalTestEncountersViral totalTestsPeopleViral totalTestsAntibody positiveTestsAntibody negativeTestsAntibody totalTestsPeopleAntibody positiveTestsPeopleAntibody negativeTestsPeopleAntibody totalTestsPeopleAntigen positiveTestsPeopleAntigen totalTestsAntigen positiveTestsAntigen fips positiveIncrease negativeIncrease total totalTestResultsSource totalTestResultsIncrease posNeg deathIncrease hospitalizedIncrease hash commercialScore negativeRegularScore negativeScore positiveScore score grade population
2020-10-12 AK 10884 0 495043 0 505927 50 0 0 0 7 0 5802 A 10/12/2020 03:59 2020-10-12T03:59:00Z 10/11 23:59 60 0 2020-10-12T03:59:00Z 505927 9748 495867 10884 60 0 0 0 0 0 0 0 0 0 0 0 0 0 2 198 2732 505927 posNeg 2930 505927 0 0 1d12a6644f8d97132e890998a17090d44e70d4e1 0 0 0 0 0 0 731545
2020-10-12 AL 166076 19853 1068387 0 1214610 856 18179 0 1892 0 1068 71240 A 10/12/2020 11:00 2020-10-12T11:00:00Z 10/12 07:00 2665 18179 2020-10-12T11:00:00Z 1214610 0 0 146223 2509 156 0 0 0 0 0 60414 0 0 0 0 0 0 1 734 3970 1234463 posNeg 4603 1234463 1 0 feb6ff2653368253706dc08cae65080d91c5d680 0 0 0 0 0 0 4903185
2020-10-12 AR 93487 4617 1066196 0 1155066 608 5960 256 0 103 737 84055 A+ 10/12/2020 00:00 2020-10-12T00:00:00Z 10/11 20:00 1586 5960 2020-10-12T00:00:00Z 1155066 0 1066196 88870 1438 148 0 0 0 0 0 0 0 0 28361 5081 21856 3300 5 654 7844 1159683 posNeg 8412 1159683 17 50 a7e93a35dd45410111c241e756412d06c15cbedf 0 0 0 0 0 0 3017804

Rhode Island (as I live in RI now)

df_states %>% filter(state == "RI") %>%
    ggplot() + 
      geom_label(x = first_day, y = 650, color = "darkgray", label = "total positive", size = 2, hjust = 0) + 
      geom_text(mapping = aes(x = date, y = 600, label = positive), color = "darkgray", size = 2, angle = 90, hjust = 0)+ 
      #geom_label(x = first_day, y = 800, color = "black", label = "death", size = 2, hjust = 0) + 
      geom_label(x = first_day, y = 550, color = tableau10[2], label = "positiveIncrease", size = 2, hjust = 0) + 
      geom_label(x = first_day, y = 500, color = tableau10[1], label = "hospitalizedCurrently", size = 2, hjust = 0) + 
      # geom_line(mapping = aes(x = date, y = death), alpha = 0.7, color = "black", size = LINE_SIZE) + 
      # geom_text(mapping = aes(x = date - 0.5, y = death + 10, label = death), color = "black", size = 1.5) + 
      # geom_point(mapping = aes(x = date, y = death), color = "black", shape = 10) + 
      geom_line(mapping = aes(x = date, y = hospitalizedCurrently), alpha = 0.7, color = tableau10[1], size = LINE_SIZE) + 
      geom_text(mapping = aes(x = date - 0.5, y = hospitalizedCurrently + 10, label = hospitalizedCurrently), color =  tableau10[1], size = 1.25) + 
      geom_point(mapping = aes(x = date, y = hospitalizedCurrently), color = tableau10[1], shape = 15) + 
      geom_line(mapping = aes(x = date, y = positiveIncrease), alpha = 0.7, color = tableau10[2], size = LINE_SIZE) + 
      geom_text(mapping = aes(x = date - 0.5, y = positiveIncrease + 10, label = positiveIncrease), color =  tableau10[2], size = 1.25)+ 
      geom_point(mapping = aes(x = date, y = positiveIncrease), color = tableau10[2]) + 
      scale_x_date(limits = c(first_day, today), breaks = seq(first_day, today, by = "day")) + 
      xlab("Date") + ylab("") + ggtitle("RI")

US - all states

df_states %>% group_by(date) %>%
    summarise(positiveIncrease = sum(positiveIncrease), hospitalizedCurrently = sum(hospitalizedCurrently), total = sum(positive)) %>% 
    ungroup() %>%
    ggplot() + 
     geom_label(x = first_day, y = 68000, color = "darkgray", label = "total positive: ", size = 2, hjust = 0) +
     geom_text(mapping = aes(x = date, y = 70000, label = total), color = "darkgray", size = 2, angle = 90, hjust = 0) +
     geom_label(x = first_day, y = 50000, color = tableau10[1], label = "hospitalizedCurrently", size = 2, hjust = 0) +
     geom_label(x = first_day, y = 55000, color = tableau10[2], label = "positiveIncrease", size = 2, hjust = 0) +
     geom_line(mapping = aes(x = date, y = hospitalizedCurrently), alpha = 0.7, color = tableau10[1], size = LINE_SIZE) +
     geom_text(mapping = aes(x = date - 0.5, y = hospitalizedCurrently + 1000, label = hospitalizedCurrently), color =  tableau10[1], size = 1.25) +
     geom_point(mapping = aes(x = date, y = hospitalizedCurrently), color = tableau10[1], shape = 15) +
     geom_line(mapping = aes(x = date, y = positiveIncrease), alpha = 0.7, color = tableau10[2], size = LINE_SIZE) +
     geom_text(mapping = aes(x = date - 0.5, y = positiveIncrease + 1000, label = positiveIncrease), color =  tableau10[2], size = 1.25) +
     geom_point(mapping = aes(x = date, y = positiveIncrease), color = tableau10[2]) +
     scale_x_date(limits = c(first_day, today), breaks = seq(first_day, today, by = "day")) +
     xlab("Date") + ylab("") + ggtitle("US - positiveIncrease & hospitalizedCurrently")

US - daily top-3 contributors

If a state has been a top 3 contributor

as_top <- df_states %>%
    filter(date > first_day)%>%
    mutate(str_date = as.character(date))%>%
    group_by(str_date) %>%
    arrange(positiveIncrease, by_group = TRUE)%>%
    slice_tail(n = 3) %>%
    ungroup() %>%
    summarise(unique(state))
as_top <- unlist(as_top)
    
df_states %>%
    filter(state %in% as_top) %>%
    ggplot() +
      stat_steamgraph(mapping = aes(x = date, y = positiveIncrease, group = state, fill = state))  +
      scale_x_date(limits = c(first_day, today), breaks = seq(first_day, today, by = "week"))  +
      scale_y_continuous(breaks = seq(-20000, 20000, by = 5000), labels = c("20000", "15000", "10000", "5000", "0", "5000", "10000", "15000", "20000")) +
      scale_fill_tableau(palette = "Tableau 20") +
      xlab("Date") + ylab("positiveIncrease") + ggtitle("If a state was a top-3 contributor on a day")

US - positiveIncrease by state

num_lag <- 21

find_coef <- function(x, y){
  m <- lm(y ~ x)
  return(coef(m)[2])
}


df_colors <-  df_states %>%
  group_by(state)%>%
  arrange(date, .by_group = TRUE) %>%
  slice_tail(n = num_lag) %>% # last N days
  summarise(trend_coef = find_coef(date, positiveIncrease)) %>% 
  mutate(trend_color = ifelse(trend_coef > 0, "increasing", ifelse(trend_coef < 0, "decreasing", "stable"))) %>% 
  ungroup()%>%
  replace(is.na(.), 0) %>%
  select(state, trend_coef, trend_color) 
 
  
df_states %>% 
    inner_join(df_colors, by = "state") %>%
    ggplot() +
      geom_smooth(mapping = aes(x = date, y = positiveIncrease), color = "gray", alpha = 0.3, method = "loess", size = LINE_SIZE) +
      geom_line(mapping = aes(x = date, y = positiveIncrease, color = trend_color), alpha = 0.7, size = LINE_SIZE) +
      geom_point(mapping = aes(x = date, y = positiveIncrease, color = trend_color), size = 1) +
      scale_x_date(limits = c(first_day, today), breaks = seq(first_day, today, by = "week")) +
      scale_colour_tableau() +
      facet_wrap(state ~ ., ncol = 6, scales = "free") +
      xlab("Date") + ylab("") + ggtitle("US - positiveIncrease by state, colored by the trend of last 21 days")

df_states %>% 
    inner_join(df_colors, by = "state") %>%
    mutate(positiveIncreasePerMillion = positiveIncrease / population * 1000000)%>%
    ggplot() +
      geom_smooth(mapping = aes(x = date, y = positiveIncreasePerMillion), color = "gray", alpha = 0.3, method = "loess", size = LINE_SIZE) +
      geom_line(mapping = aes(x = date, y = positiveIncreasePerMillion, color = trend_color), alpha = 0.7, size = LINE_SIZE) +
      geom_point(mapping = aes(x = date, y = positiveIncreasePerMillion, color = trend_color), size = 1) +
      scale_y_continuous(limits = c(0, 600), breaks = seq(0, 600, by = 150)) +
      scale_x_date(limits = c(first_day, today), breaks = seq(first_day, today, by = "week")) +
      scale_colour_tableau() +
      facet_wrap(state ~ ., ncol = 6, scales = "free")  +
      xlab("Date") + ylab("") + ggtitle("US - positiveIncreasePerMillion by state, colored by the trend of last 21 days")

US - hospitalizedCurrently by state

df_states %>% 
    ggplot() +
      geom_smooth(mapping = aes(x = date, y = hospitalizedCurrently), color = "gray", alpha = 0.3, method = "loess", size = LINE_SIZE) +
      geom_line(mapping = aes(x = date, y = hospitalizedCurrently), alpha = 0.7, color = tableau10[3], size = LINE_SIZE) +
      geom_point(mapping = aes(x = date, y = hospitalizedCurrently), color = tableau10[3], size = 1) +
      scale_x_date(limits = c(first_day, today), breaks = seq(first_day, today, by = "week")) +
      facet_wrap(state ~ ., ncol = 6, scales = "free") +
      xlab("Date") + ylab("") + ggtitle("US - hospitalizedCurrently by state")

US - dailyTestPositiveRate against overallTestedPopulationRate

df_pr <- df_states %>% 
    mutate(testPositiveRate = positiveIncrease / totalTestResultsIncrease, testedPopulationRate = totalTestResults / population) %>%
    filter(testPositiveRate > 0 & testPositiveRate < 1) # rm buggy data to allow log scales
  
df_pr_colors <-  df_pr %>%
  group_by(state)%>%
  arrange(date, .by_group = TRUE) %>%
  slice_tail(n = num_lag) %>% # last N days
  summarise(trend_coef = find_coef(date, testPositiveRate)) %>% 
  mutate(trend_color = ifelse(trend_coef > 0, "increasing", ifelse(trend_coef < 0, "decreasing", "stable"))) %>% 
  ungroup()%>%
  replace(is.na(.), 0) %>%
  select(state, trend_coef, trend_color) 

df_pr %>%
 inner_join(df_pr_colors, by = "state") %>%
 ggplot() +
    geom_smooth(mapping = aes(x = testedPopulationRate, y = testPositiveRate), color = "gray", alpha = 0.3, method = "loess", size = LINE_SIZE) +
    geom_line(mapping = aes(x = testedPopulationRate, y = testPositiveRate, color = trend_color), alpha = 0.7, size = LINE_SIZE) +
    geom_point(mapping = aes(x = testedPopulationRate, y = testPositiveRate, color = trend_color), size = 1) +
    scale_x_continuous(limits = c(0, 1.0), breaks = seq(0, 1.0, by = 0.05)) +
    scale_y_continuous(limits = c(0.001, 1), trans = 'log10', breaks = c(0.001, 0.01, 0.05, 0.1, 0.2, 0.3, 0.5, 0.75, 1)) +
    scale_colour_tableau() +
    facet_wrap(state ~ ., ncol = 6, scales = "free")  +
    xlab("dailyTestPositiveRate") + ylab("overallTestedPopulationRate") + ggtitle("US - dailyTestPositiveRate against overallTestedPopulationRate")

US - death per 10k by state

df_states %>% 
    mutate(deathPer10K = death / population * 10000) %>%
    ggplot() +
     geom_line(mapping = aes(x = date, y = deathPer10K), alpha = 0.7, color = tableau10[3], size = LINE_SIZE) +
     geom_point(mapping = aes(x = date, y = deathPer10K), color = tableau10[3], size = 1) +
     scale_x_date(limits = c(first_day, today), breaks = seq(first_day, today, by = "week")) +
     scale_y_continuous(limits = c(0, 20), breaks = seq(0, 20, by = 5)) +
     facet_wrap(state ~ ., ncol = 6, scales = "free")  +
     xlab("date") + ylab("death per 10k") + ggtitle("US - death per 10k by state")

US - positive per 1k by state

df_states %>% 
    mutate(positivePerOneK = positive / population * 1000) %>%
    ggplot() +
      geom_line(mapping = aes(x = date, y = positivePerOneK), alpha = 0.7, color = tableau10[4], size = LINE_SIZE) +
      geom_point(mapping = aes(x = date, y = positivePerOneK), color = tableau10[4], size = 1) +
      scale_y_continuous(limits = c(0, 25), breaks = seq(0, 25, by = 5)) +
      scale_x_date(limits = c(first_day, today), breaks = seq(first_day, today, by = "week")) +
      facet_wrap(state ~ ., ncol = 6, scales = "free") +
      xlab("date") + ylab("") + ggtitle("US - positivePerOneK by state")

US - tested amount by state

df_states %>% 
    mutate(testResultsIncrease = positiveIncrease + negativeIncrease) %>%
    ggplot() +
      geom_smooth(mapping = aes(x = date, y = testResultsIncrease), color = "gray", alpha = 0.3, method = "loess", size = LINE_SIZE) +
      geom_line(mapping = aes(x = date, y = testResultsIncrease), alpha = 0.7, color = tableau10[7], size = LINE_SIZE) +
      geom_point(mapping = aes(x = date, y = testResultsIncrease), color = tableau10[7], size = 1) +
      scale_x_date(limits = c(first_day, today), breaks = seq(first_day, today, by = "week")) +
      facet_wrap(state ~ ., ncol = 6, scales = "free")  +
      xlab("date") + ylab("testResultsIncrease") + ggtitle("US - testResultsIncrease by state")